ImageBench

ImageBench V1 —

192 evaluations across 6 categories

Benchmark V1 verdicts are produced by VLM judges and can contain mistakes. Treat PASS/FAIL labels as machine-assisted assessments, and inspect the images yourself. Learn more about the methodology.

Generation Details

Source-backed model context, size, cost, and request settings for this ImageBench V1 run.

fal/luma/uni-1-max

API

UNI-1 Max is Luma AI's flagship of its Uni-1 ('Unified Intelligence') line — a decoder-only autoregressive multimodal model that reasons before generating, positioned as higher quality than Luma's earlier Photon. It measures around ELO ~1175 on the image arena, among the strongest text-to-image models, with reported quality rivaling Google and OpenAI at lower cost.

Maker
Luma AI
Family
UNI-1 Max
Model Size
not disclosed
unknown
Cost
~$0.102 / image
medium
Run Target
fal/luma/agent/uni-1/v1/max
Effective Request
aspect_ratio: 1:1
Sources
Luma UNI-1 Max on falLuma opens Uni-1.1 APILuma UNI-1
62.0
Overall
79%
Capability
45.5
Est. Preference
151
Pass
41
Fail
88.2s
Avg Latency
63.0s
Min Latency
179.0s
Max Latency
Text Rendering100%Spatial Reasoning91%Human realism55%Truthfulness48%Professional Studio93%Graphical design96%Preference46%Latency0%

All 192 generations

Text Rendering100%

Typography Style100%

Writing accuracy100%

Spatial Reasoning91%

Attributes Binding100%

Compositionality100%

Counting89%

Negation100%

Relative Position83%

Scale & Proportions78%

Human realism55%

Faces & Expressions75%

Full Body17%

Hands83%

Multi-Subject33%

Truthfulness48%

Photorealism67%

Physics & Reflections50%

World Knowledge42%

Professional Studio93%

Camera & Lighting92%

Color Precision92%

Photorealism100%

Graphical design96%

Data Visualisation100%

Layout & Design89%

Style Diversity100%